GitHub radar
Loop Engineering: Design Systems That Prompt Your AI Agents
Loop Engineering is a pattern library & CLI toolkit for replacing manual prompting of AI code agents by building a system which automatically prompts the agent periodically
Loop Engineering is a collection of patterns, templates & CLI tools for building systems which orchestrate AI code agents such as Claude Code, Codex, Cursor, etc. The idea behind Loop Engineering is to “build a loop” rather than manually interacting with the code agent. A loop interacts with the code agent periodically via persistent state, budgets & skill files. With the help of the loop init CLI tool, you can spin up a new loop in ~5 mins by creating STATE.md / LOOP.md files, budget/constraint files and obtain a Loop Ready Score. You can use the loop doctor CLI tool to find out the top 3 things you can do next. There are other CLI tools such as loop-audit (to obtain your Loop Ready Score), loop-cost (to estimate your token expenditure), loop-sync (to detect drift), loop-context (to manage memory), loop-worktree (to create git worktrees), loop-gate (to enforce path denylists) & an MCP server (to lookup patterns). According to Boris Cherny (Head of Claude Code @ Anthropic): “I don’t prompt Claude anymore. I have loops that prompt Claude and figure out what to do.” Loop Engineering was launched only 2 months back and already has over 10,500+ stars.
Why a vibe-coder should care
Loop Engineering helps you create structural blueprints for building autonomous agent workflows if you are someone who frequently uses AI code agents. With the help of the Loop Ready Score, you can measure the autonomy of your loop and decide whether it requires manual intervention or not.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up loop-engineering for this project: https://github.com/cobusgreyling/loop-engineering — run npx @cobusgreyling/loop init . --pattern daily-triage --tool claude, then npx @cobusgreyling/loop doctor . and show me the Loop Ready Score and top 3 actions to improve it.
Runs on a regular laptop — only Node.js needed.
Open on GitHub▌ More finds